Abstract
This paper describes a low-cost algorithm for tracking lane boundaries in a sequence of images. The algorithm is destined to painted road with slow curvature. The basic idea proposed in our approach is that complete processing of each image can be avoided using the knowledge of the lane markings position in the previous ones. The markings detection is obtained using Radon transform that exploits the markings brilliance relatively to the road surface. The experimental tests proved the robustness of this approach even in shadows presence. The originality of our approach compared to those using the Hough transform is that it does not require any tresholding step and edge detection operator. Consequently, the proposed method is proved to be much faster.